Prev: Which CPU or core to disable?
Next: 3 pages printed from Solaris 8 before the actual print job
From: murali on 12 Jun 2010 00:51
Could some one let me know the easy way of understanding solaris disk
naming conventions. It is bit difficult to understand. I know in
linux it is named /dev/hda or /dev/sda(for scsi). Some one advise on
this.
From: Doug McIntyre on 12 Jun 2010 01:19
murali <murali.krishnaa7(a)gmail.com> writes:
>Could some one let me know the easy way of understanding solaris disk
>naming conventions. It is bit difficult to understand. I know in
>linux it is named /dev/hda or /dev/sda(for scsi). Some one advise on
>this.

Its usually based on controller number, target ID number, LUN, and slice
numbers. It makes it alot easier to find disks when you have more than
a couple of them. Ie. my newest system has 50 drives online. (Granted,
the SAS expanders don't exactly make it too easy to find the mapping
for which disk enclosure and port number map to which disk).

Ie. /dev/dsk/c0t0d0s0
Or /dev/dsk/c3t23d0s2

Controller 3, target 23, lun0, slice 2.

FC WWN's make the format alot worse, since the disk number is usually
including the WWN at that point.

But, once you move past the PC notion of having only 4 disks or so,
its quite a bit easier to track down bad disks or so when you need to,
figuring out which disk shelf is attached to which controller, which
drive in that enclosure, etc.


From: Andrew Gabriel on 12 Jun 2010 15:11
In article <4c1318c9$0$50146$8046368a(a)newsreader.iphouse.net>,
Doug McIntyre <merlyn(a)geeks.org> writes:
> murali <murali.krishnaa7(a)gmail.com> writes:
>>Could some one let me know the easy way of understanding solaris disk
>>naming conventions. It is bit difficult to understand. I know in
>>linux it is named /dev/hda or /dev/sda(for scsi). Some one advise on
>>this.
>
> Its usually based on controller number, target ID number, LUN, and slice
> numbers. It makes it alot easier to find disks when you have more than
> a couple of them. Ie. my newest system has 50 drives online. (Granted,
> the SAS expanders don't exactly make it too easy to find the mapping
> for which disk enclosure and port number map to which disk).
>
> Ie. /dev/dsk/c0t0d0s0
> Or /dev/dsk/c3t23d0s2
>
> Controller 3, target 23, lun0, slice 2.
>
> FC WWN's make the format alot worse, since the disk number is usually
> including the WWN at that point.
>
> But, once you move past the PC notion of having only 4 disks or so,
> its quite a bit easier to track down bad disks or so when you need to,
> figuring out which disk shelf is attached to which controller, which
> drive in that enclosure, etc.

Just to add a bit more...

On Solaris x86 with IDE disks, sometimes the target is omitted from
the device name (it depends which driver is driving the controller),
so you might have:

/dev/dsk/c0d0s0 primary master
/dev/dsk/c0d1s0 primary slave
/dev/dsk/c1d0s0 secondary master
/dev/dsk/c1d1s0 secondary slave

Another thing worth pointing out is that controller numbers are
persistent in Solaris, so controller numbers (and hence device
names) never change when you add or remove controllers or disks.

Slices and partitions are identified by the trailing s0 or p0 on
device names; s are Solaris slices, and p are FDISK partitions.
(Unfortunately, in many places Solaris slices are also confusingly
called partitions.)
There are two types of partioning/slicing for disks.

GPT/EFI labels. This is an industry standard partitioning scheme,
originally developed by Intel for Itanium but is expected to take
over from FDISK partitioning in due course. In Solaris, large disks
must be EFI labelled, but you can't boot from EFI labelled disks
yet. EFI labels are architecture neutral, so they work on both sparc
and x86. Solaris supports up to 9 EFI partitions, 0-8.

SMI labels. This is the more traditional partitioning scheme on
Solaris (and going back into older SunOS releases too). It works
rather differently on sparc and x86. Boot disks must be SMI labeled.

On sparc, there are up to 8 slices, s0-s8. By convention, slice 2
(s2) is the whole disk, and thus overlaps all other slices which
are in use, although it is possible to create s2 differently from
this.

On x86, this slice structure exists inside the SOLARIS primary
FDISK partition (and s2 is this case is the whole of the SOLARIS
primary FDISK partition). Solaris x86 supports 16 slices, verses
Solaris sparc's 8 slices. (SVR4 supported 16 slices which is where
the 16 on Solaris x86 come from, but sparc had to remain backwards
compatible with the openboot prom, and was thus limited to 8.)
Additional devices are provided to allow addressing of FDISK
partitions: p0 is the whole disk irrespective of any partitioning
on the disk (which is something sparc is missing), and p1-p4 are
the 4 primary fdisk partitions. s0-s15 are the 16 slices inside
the SOLARIS FDISK partition. There has been some recent work to
provide device access to logical disks inside an Extended DOS
FDISK partition. I think they use p5 and up (like SysV/386 did),
but I haven't played with this yet.

--
Andrew Gabriel
[email address is not usable -- followup in the newsgroup]
From: Barry on 23 Jun 2010 02:58
On 12 Jun 2010 05:19:05 GMT, Doug McIntyre <merlyn(a)geeks.org> wrote:

>murali <murali.krishnaa7(a)gmail.com> writes:
>>Could some one let me know the easy way of understanding solaris disk
>>naming conventions. It is bit difficult to understand. I know in
>>linux it is named /dev/hda or /dev/sda(for scsi). Some one advise on
>>this.
>
>Its usually based on controller number, target ID number, LUN, and slice
>numbers. It makes it alot easier to find disks when you have more than
>a couple of them. Ie. my newest system has 50 drives online. (Granted,
>the SAS expanders don't exactly make it too easy to find the mapping
>for which disk enclosure and port number map to which disk).
>
>Ie. /dev/dsk/c0t0d0s0
>Or /dev/dsk/c3t23d0s2
>
>Controller 3, target 23, lun0, slice 2.
>
>FC WWN's make the format alot worse, since the disk number is usually
>including the WWN at that point.

By alot do you mean not a lot in the way atheist means not a theist?

>But, once you move past the PC notion of having only 4 disks or so,
>its quite a bit easier to track down bad disks or so when you need to,
>figuring out which disk shelf is attached to which controller, which
>drive in that enclosure, etc.

 | 
Pages: 1
Prev: Which CPU or core to disable?
Next: 3 pages printed from Solaris 8 before the actual print job